The $N$-1-1 contingency reliability criterion considers the consecutive loss of two components in a power system, with intervening time for system adjustments between the two losses. In this paper, we consider the problem of optimizing generation unit while ensuring the $N$-1-1 criterion. Due to the coupling of time periods associated with consecutive component losses, the resulting problem yields a very large-scale mixed-integer linear optimization model. For efficient solution, we introduce a novel branch-and-cut algorithm using a temporally decomposed bilevel separation oracle. The model and algorithm are assessed using multiple IEEE test systems, and a comprehensive analysis is performed to compare system performance across different contingency criteria. Computational results demonstrate the value of considering intervening time for system adjustments in terms of total cost and system robustness.